├── .gitignore ├── LICENSE ├── Makefile ├── Makefile.feature.esb1 ├── Makefile.feature.j1 ├── Makefile.feature.n1 ├── Makefile.krs1 ├── Makefile.lgb1 ├── Makefile.lgb2 ├── README.md ├── notebook └── README.md ├── requirements.txt ├── src ├── const.py ├── create_fmap_esb.py ├── evaluate.py ├── generate_j1.py ├── generate_n1.py ├── train_predict_krs1.py └── train_predict_lgb1.py └── tox.ini /.gitignore: -------------------------------------------------------------------------------- 1 | # Byte-compiled / optimized / DLL files 2 | __pycache__/ 3 | *.py[cod] 4 | *$py.class 5 | 6 | # C extensions 7 | *.so 8 | 9 | # Distribution / packaging 10 | .Python 11 | env/ 12 | build/ 13 | develop-eggs/ 14 | dist/ 15 | downloads/ 16 | eggs/ 17 | .eggs/ 18 | lib/ 19 | lib64/ 20 | parts/ 21 | sdist/ 22 | var/ 23 | *.egg-info/ 24 | .installed.cfg 25 | *.egg 26 | 27 | # PyInstaller 28 | # Usually these files are written by a python script from a template 29 | # before PyInstaller builds the exe, so as to inject date/other infos into it. 30 | *.manifest 31 | *.spec 32 | 33 | # Installer logs 34 | pip-log.txt 35 | pip-delete-this-directory.txt 36 | 37 | # Unit test / coverage reports 38 | htmlcov/ 39 | .tox/ 40 | .coverage 41 | .coverage.* 42 | .cache 43 | nosetests.xml 44 | coverage.xml 45 | *,cover 46 | .hypothesis/ 47 | 48 | # Translations 49 | *.mo 50 | *.pot 51 | 52 | # Django stuff: 53 | *.log 54 | local_settings.py 55 | 56 | # Flask stuff: 57 | instance/ 58 | .webassets-cache 59 | 60 | # Scrapy stuff: 61 | .scrapy 62 | 63 | # Sphinx documentation 64 | docs/_build/ 65 | 66 | # PyBuilder 67 | target/ 68 | 69 | # IPython Notebook 70 | .ipynb_checkpoints 71 | 72 | # pyenv 73 | .python-version 74 | 75 | # celery beat schedule file 76 | celerybeat-schedule 77 | 78 | # dotenv 79 | .env 80 | 81 | # virtualenv 82 | venv/ 83 | ENV/ 84 | 85 | # Spyder project settings 86 | .spyderproject 87 | 88 | # Rope project settings 89 | .ropeproject 90 | 91 | # Kaggle 92 | input/ 93 | -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | {one line to give the program's name and a brief idea of what it does.} 635 | Copyright (C) {year} {name of author} 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | {project} Copyright (C) {year} {fullname} 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /Makefile: -------------------------------------------------------------------------------- 1 | # XXX: competition name 2 | COMPETITION := cat-in-the-dat-ii 3 | 4 | # gsed on macOS. sed on LINUX 5 | SED := gsed 6 | 7 | # directories 8 | DIR_DATA := input 9 | DIR_BUILD := build 10 | DIR_FEATURE := $(DIR_BUILD)/feature 11 | DIR_METRIC := $(DIR_BUILD)/metric 12 | DIR_MODEL := $(DIR_BUILD)/model 13 | 14 | # directories for the cross validation and ensembling 15 | DIR_VAL := $(DIR_BUILD)/val 16 | DIR_TST := $(DIR_BUILD)/tst 17 | DIR_SUB := $(DIR_BUILD)/sub 18 | 19 | DIRS := $(DIR_DATA) $(DIR_BUILD) $(DIR_FEATURE) $(DIR_METRIC) $(DIR_MODEL) \ 20 | $(DIR_VAL) $(DIR_TST) $(DIR_SUB) 21 | 22 | # data files for training and predict 23 | DATA_TRN := $(DIR_DATA)/train.csv 24 | DATA_TST := $(DIR_DATA)/test.csv 25 | SAMPLE_SUBMISSION := $(DIR_DATA)/sample_submission.csv 26 | 27 | LABEL_IDX = 25 28 | 29 | ID_TST := $(DIR_DATA)/id.tst.csv 30 | HEADER := $(DIR_DATA)/header.csv 31 | 32 | Y_TRN:= $(DIR_FEATURE)/y.trn.txt 33 | Y_TST:= $(DIR_FEATURE)/y.tst.txt 34 | 35 | data: $(DATA_TRN) $(DATA_TST) $(SAMPLE_SUBMISSION) 36 | 37 | $(DIRS): 38 | mkdir -p $@ 39 | 40 | $(DATA_TRN) $(DATA_TST) $(SAMPLE_SUBMISSION): | $(DIR_DATA) 41 | kaggle competitions download -c $(COMPETITION) -p $(DIR_DATA) 42 | find . -name "*.zip" -exec sh -c 'unzip -d `dirname {}` {}' ';' 43 | 44 | $(HEADER): $(SAMPLE_SUBMISSION) 45 | head -1 $< > $@ 46 | 47 | $(ID_TST): $(SAMPLE_SUBMISSION) 48 | cut -d, -f1 $< | tail -n +2 > $@ 49 | 50 | $(Y_TST): $(SAMPLE_SUBMISSION) | $(DIR_FEATURE) 51 | cut -d, -f2 $< | tail -n +2 > $@ 52 | 53 | $(Y_TRN): $(DATA_TRN) | $(DIR_FEATURE) 54 | cut -d, -f$(LABEL_IDX) $< | tail -n +2 > $@ 55 | 56 | # cleanup 57 | clean:: 58 | find . -name '*.pyc' -delete 59 | 60 | clobber: clean 61 | -rm -rf $(DIR_DATA) $(DIR_BUILD) 62 | 63 | .PHONY: clean clobber mac.setup ubuntu.setup apt.setup pip.setup 64 | -------------------------------------------------------------------------------- /Makefile.feature.esb1: -------------------------------------------------------------------------------- 1 | include Makefile 2 | 3 | FEATURE_NAME := esb1 4 | 5 | BASE_MODELS := lgb1_j1 \ 6 | lgb2_j1 7 | 8 | PREDICTS_TRN := $(foreach m, $(BASE_MODELS), $(DIR_VAL)/$(m).val.yht) 9 | PREDICTS_TST := $(foreach m, $(BASE_MODELS), $(DIR_TST)/$(m).tst.yht) 10 | 11 | FEATURE_TRN := $(DIR_FEATURE)/$(FEATURE_NAME).trn.csv 12 | FEATURE_TST := $(DIR_FEATURE)/$(FEATURE_NAME).tst.csv 13 | FEATURE_MAP := $(DIR_FEATURE)/$(FEATURE_NAME).fmap 14 | 15 | $(FEATURE_MAP): | $(DIR_FEATURE) 16 | python src/create_fmap_esb.py --base-models $(BASE_MODELS) \ 17 | --feature-map-file $@ 18 | 19 | $(FEATURE_TRN): $(Y_TRN) $(PREDICTS_TRN) | $(DIR_FEATURE) 20 | paste -d, $^ | tr -d '\r' > $@ 21 | 22 | $(FEATURE_TST): $(Y_TST) $(PREDICTS_TST) | $(DIR_FEATURE) 23 | paste -d, $^ | tr -d '\r' > $@ 24 | 25 | 26 | clean:: clean_$(FEATURE_NAME) 27 | 28 | clean_$(FEATURE_NAME): 29 | -rm $(FEATURE_TRN) $(FEATURE_TST) 30 | -------------------------------------------------------------------------------- /Makefile.feature.j1: -------------------------------------------------------------------------------- 1 | #-------------------------------------------------------------------------- 2 | # j1: features for xgboost, lightgbm, rf, et 3 | #-------------------------------------------------------------------------- 4 | include Makefile 5 | 6 | FEATURE_NAME := j1 7 | 8 | FEATURE_TRN := $(DIR_FEATURE)/$(FEATURE_NAME).trn.h5 9 | FEATURE_TST := $(DIR_FEATURE)/$(FEATURE_NAME).tst.h5 10 | FEATURE_MAP := $(DIR_FEATURE)/$(FEATURE_NAME).fmap 11 | 12 | $(FEATURE_TRN) $(FEATURE_TST) $(FEATURE_MAP): $(DATA_TRN) $(DATA_TST) | $(DIR_FEATURE) 13 | python ./src/generate_$(FEATURE_NAME).py --train-file $< \ 14 | --test-file $(lastword $^) \ 15 | --train-feature-file $(FEATURE_TRN) \ 16 | --test-feature-file $(FEATURE_TST) \ 17 | --feature-map-file $(FEATURE_MAP) 18 | -------------------------------------------------------------------------------- /Makefile.feature.n1: -------------------------------------------------------------------------------- 1 | #-------------------------------------------------------------------------- 2 | # n1: features for keras, lr, fm 3 | #-------------------------------------------------------------------------- 4 | include Makefile 5 | 6 | FEATURE_NAME := n1 7 | 8 | FEATURE_TRN := $(DIR_FEATURE)/$(FEATURE_NAME).trn.sps 9 | FEATURE_TST := $(DIR_FEATURE)/$(FEATURE_NAME).tst.sps 10 | FEATURE_MAP := $(DIR_FEATURE)/$(FEATURE_NAME).fmap 11 | 12 | $(FEATURE_TRN) $(FEATURE_TST) $(FEATURE_MAP): $(DATA_TRN) $(DATA_TST) | $(DIR_FEATURE) 13 | python ./src/generate_$(FEATURE_NAME).py --train-file $< \ 14 | --test-file $(lastword $^) \ 15 | --train-feature-file $(FEATURE_TRN) \ 16 | --test-feature-file $(FEATURE_TST) \ 17 | --feature-map-file $(FEATURE_MAP) 18 | -------------------------------------------------------------------------------- /Makefile.krs1: -------------------------------------------------------------------------------- 1 | include Makefile.feature.n1 2 | 3 | N = 55 4 | ALGO_NAME := krs1 5 | MODEL_NAME := $(ALGO_NAME)_$(FEATURE_NAME) 6 | 7 | METRIC_VAL := $(DIR_METRIC)/$(MODEL_NAME).val.txt 8 | 9 | PREDICT_VAL := $(DIR_VAL)/$(MODEL_NAME).val.yht 10 | PREDICT_TST := $(DIR_TST)/$(MODEL_NAME).tst.yht 11 | 12 | SUBMISSION_TST := $(DIR_SUB)/$(MODEL_NAME).sub.csv 13 | SUBMISSION_TST_GZ := $(DIR_SUB)/$(MODEL_NAME).sub.csv.gz 14 | 15 | all: validation submission 16 | validation: $(METRIC_VAL) 17 | submission: $(SUBMISSION_TST) 18 | retrain: clean_$(ALGO_NAME) submission 19 | 20 | submit: $(SUBMISSION_TST) 21 | kaggle competitions submit -c $(COMPETITION) -f $< -m $(MODEL_NAME) 22 | 23 | $(PREDICT_TST) $(PREDICT_VAL): $(FEATURE_TRN) $(FEATURE_TST) \ 24 | | $(DIR_VAL) $(DIR_TST) 25 | python ./src/train_predict_krs1.py --train-file $< \ 26 | --test-file $(word 2, $^) \ 27 | --predict-valid-file $(PREDICT_VAL) \ 28 | --predict-test-file $(PREDICT_TST) \ 29 | --n-est $(N) 30 | 31 | $(SUBMISSION_TST_GZ): $(SUBMISSION_TST) 32 | gzip $< 33 | 34 | $(SUBMISSION_TST): $(PREDICT_TST) $(HEADER) $(ID_TST) | $(DIR_SUB) 35 | paste -d, $(lastword $^) $< > $@.tmp 36 | cat $(word 2, $^) $@.tmp > $@ 37 | rm $@.tmp 38 | 39 | $(METRIC_VAL): $(PREDICT_VAL) $(Y_TRN) | $(DIR_METRIC) 40 | python ./src/evaluate.py --predict-file $< \ 41 | --target-file $(lastword $^) > $@ 42 | cat $@ 43 | 44 | 45 | clean:: clean_$(ALGO_NAME) 46 | 47 | clean_$(ALGO_NAME): 48 | -rm $(METRIC_VAL) $(PREDICT_VAL) $(PREDICT_TST) $(SUBMISSION_TST) 49 | find . -name '*.pyc' -delete 50 | 51 | .DEFAULT_GOAL := all 52 | -------------------------------------------------------------------------------- /Makefile.lgb1: -------------------------------------------------------------------------------- 1 | include Makefile.feature.j1 2 | 3 | N = 10000 4 | N_LEAF = 200 5 | LRATE = 0.1 6 | N_MIN = 10 7 | SUBCOL = 0.8 8 | SUBROW = 0.7 9 | SUBROW_FREQ = 1 10 | N_STOP = 20 11 | ALGO_NAME := lgb1 12 | MODEL_NAME := $(ALGO_NAME)_$(FEATURE_NAME) 13 | 14 | METRIC_VAL := $(DIR_METRIC)/$(MODEL_NAME).val.txt 15 | 16 | PREDICT_VAL := $(DIR_VAL)/$(MODEL_NAME).val.yht 17 | PREDICT_TST := $(DIR_TST)/$(MODEL_NAME).tst.yht 18 | FEATURE_IMP := $(DIR_MODEL)/$(MODEL_NAME).imp.csv 19 | 20 | SUBMISSION_TST := $(DIR_SUB)/$(MODEL_NAME).sub.csv 21 | SUBMISSION_TST_GZ := $(DIR_SUB)/$(MODEL_NAME).sub.csv.gz 22 | 23 | all: validation submission 24 | validation: $(METRIC_VAL) 25 | submission: $(SUBMISSION_TST) 26 | retrain: clean_$(ALGO_NAME) submission 27 | 28 | submit: $(SUBMISSION_TST) 29 | kaggle competitions submit -c $(COMPETITION) -f $< -m $(MODEL_NAME) 30 | 31 | $(PREDICT_TST) $(PREDICT_VAL): $(FEATURE_TRN) $(FEATURE_TST) \ 32 | | $(DIR_VAL) $(DIR_TST) 33 | ./src/train_predict_lgb1.py --train-file $< \ 34 | --test-file $(word 2, $^) \ 35 | --predict-valid-file $(PREDICT_VAL) \ 36 | --predict-test-file $(PREDICT_TST) \ 37 | --n-est $(N) \ 38 | --n-leaf $(N_LEAF) \ 39 | --lrate $(LRATE) \ 40 | --n-min $(N_MIN) \ 41 | --subcol $(SUBCOL) \ 42 | --subrow $(SUBROW) \ 43 | --subrow-freq $(SUBROW_FREQ) \ 44 | --early-stop $(N_STOP) 45 | 46 | $(SUBMISSION_TST_GZ): $(SUBMISSION_TST) 47 | gzip $< 48 | 49 | $(SUBMISSION_TST): $(PREDICT_TST) $(HEADER) $(ID_TST) | $(DIR_SUB) 50 | paste -d, $(lastword $^) $< > $@.tmp 51 | cat $(word 2, $^) $@.tmp > $@ 52 | rm $@.tmp 53 | 54 | $(METRIC_VAL): $(PREDICT_VAL) $(Y_TRN) | $(DIR_METRIC) 55 | python ./src/evaluate.py --predict-file $< \ 56 | --target-file $(lastword $^) > $@ 57 | cat $@ 58 | 59 | 60 | clean:: clean_$(ALGO_NAME) 61 | 62 | clean_$(ALGO_NAME): 63 | -rm $(METRIC_VAL) $(PREDICT_VAL) $(PREDICT_TST) $(SUBMISSION_TST) 64 | find . -name '*.pyc' -delete 65 | 66 | .DEFAULT_GOAL := all 67 | -------------------------------------------------------------------------------- /Makefile.lgb2: -------------------------------------------------------------------------------- 1 | include Makefile.feature.j1 2 | 3 | N = 10000 4 | N_LEAF = 128 5 | LRATE = 0.05 6 | N_MIN = 25 7 | SUBCOL = 0.5 8 | SUBROW = 0.7 9 | SUBROW_FREQ = 1 10 | N_STOP = 20 11 | ALGO_NAME := lgb2 12 | MODEL_NAME := $(ALGO_NAME)_$(FEATURE_NAME) 13 | 14 | METRIC_VAL := $(DIR_METRIC)/$(MODEL_NAME).val.txt 15 | 16 | PREDICT_VAL := $(DIR_VAL)/$(MODEL_NAME).val.yht 17 | PREDICT_TST := $(DIR_TST)/$(MODEL_NAME).tst.yht 18 | FEATURE_IMP := $(DIR_MODEL)/$(MODEL_NAME).imp.csv 19 | 20 | SUBMISSION_TST := $(DIR_SUB)/$(MODEL_NAME).sub.csv 21 | SUBMISSION_TST_GZ := $(DIR_SUB)/$(MODEL_NAME).sub.csv.gz 22 | 23 | all: validation submission 24 | validation: $(METRIC_VAL) 25 | submission: $(SUBMISSION_TST) 26 | retrain: clean_$(ALGO_NAME) submission 27 | 28 | submit: $(SUBMISSION_TST) 29 | kaggle competitions submit -c $(COMPETITION) -f $< -m $(MODEL_NAME) 30 | 31 | $(PREDICT_TST) $(PREDICT_VAL): $(FEATURE_TRN) $(FEATURE_TST) \ 32 | | $(DIR_VAL) $(DIR_TST) 33 | ./src/train_predict_lgb1.py --train-file $< \ 34 | --test-file $(word 2, $^) \ 35 | --predict-valid-file $(PREDICT_VAL) \ 36 | --predict-test-file $(PREDICT_TST) \ 37 | --n-est $(N) \ 38 | --n-leaf $(N_LEAF) \ 39 | --lrate $(LRATE) \ 40 | --n-min $(N_MIN) \ 41 | --subcol $(SUBCOL) \ 42 | --subrow $(SUBROW) \ 43 | --subrow-freq $(SUBROW_FREQ) \ 44 | --early-stop $(N_STOP) 45 | 46 | $(SUBMISSION_TST_GZ): $(SUBMISSION_TST) 47 | gzip $< 48 | 49 | $(SUBMISSION_TST): $(PREDICT_TST) $(HEADER) $(ID_TST) | $(DIR_SUB) 50 | paste -d, $(lastword $^) $< > $@.tmp 51 | cat $(word 2, $^) $@.tmp > $@ 52 | rm $@.tmp 53 | 54 | $(METRIC_VAL): $(PREDICT_VAL) $(Y_TRN) | $(DIR_METRIC) 55 | python ./src/evaluate.py --predict-file $< \ 56 | --target-file $(lastword $^) > $@ 57 | cat $@ 58 | 59 | 60 | clean:: clean_$(ALGO_NAME) 61 | 62 | clean_$(ALGO_NAME): 63 | -rm $(METRIC_VAL) $(PREDICT_VAL) $(PREDICT_TST) $(SUBMISSION_TST) 64 | find . -name '*.pyc' -delete 65 | 66 | .DEFAULT_GOAL := all 67 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Kaggler TV Episode #4: Kaggle Competition Pipeline Demo 2 | 3 | Demo video is available on [YouTube](https://youtu.be/861NAO5-XJo) 4 | 5 | - Comments are available in the subtitle. Please turn on the subtitle. 6 | - Kaggler pipeline template: https://github.com/jeongyoonlee/kaggler-template 7 | - GitHub repo used in the demo: https://github.com/jeongyoonlee/cat-in-the-dat-ii 8 | - Blog post about the pipeline: http://kaggler.com/2015/09/21/kagglers-toolbox.html 9 | 10 | Made by Kagglers, for Kagglers. 11 | 12 | Kaggler TV uploads one video a week every Sunday at 10am Pacific Time. 13 | 14 | You can find the schedule and request for contents at Kaggler TV GitHub: 15 | https://github.com/kaggler-tv/kaggler-tv-schedule 16 | -------------------------------------------------------------------------------- /notebook/README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/jeongyoonlee/kaggler-template/eb30077e28ae40db9639c1125ce479183441c35d/notebook/README.md -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- 1 | numpy 2 | scipy 3 | pandas 4 | lightgbm 5 | keras 6 | cython 7 | h5py 8 | kaggle 9 | kaggler 10 | -------------------------------------------------------------------------------- /src/const.py: -------------------------------------------------------------------------------- 1 | TARGET_COL = 'target' 2 | ID_COL = 'id' 3 | N_FOLD = 5 4 | SEED = 42 5 | -------------------------------------------------------------------------------- /src/create_fmap_esb.py: -------------------------------------------------------------------------------- 1 | import argparse 2 | 3 | 4 | if __name__ == '__main__': 5 | parser = argparse.ArgumentParser() 6 | parser.add_argument('--base-models', required=True, nargs='+', 7 | dest='base_models') 8 | parser.add_argument('--feature-map-file', required=True, 9 | dest='feature_map_file') 10 | 11 | args = parser.parse_args() 12 | 13 | with open(args.feature_map_file, 'w') as f: 14 | for i, col in enumerate(args.base_models): 15 | f.write('{}\t{}\tq\n'.format(i, col)) 16 | 17 | -------------------------------------------------------------------------------- /src/evaluate.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | from sklearn.metrics import roc_auc_score as AUC 3 | import argparse 4 | import json 5 | import numpy as np 6 | import os 7 | 8 | 9 | if __name__ == '__main__': 10 | 11 | parser = argparse.ArgumentParser() 12 | parser.add_argument('--target-file', '-t', required=True, dest='target_file') 13 | parser.add_argument('--predict-file', '-p', required=True, dest='predict_file') 14 | 15 | args = parser.parse_args() 16 | 17 | p = np.loadtxt(args.predict_file, delimiter=',') 18 | y = np.loadtxt(args.target_file, delimiter=',') 19 | 20 | model_name = os.path.splitext(os.path.splitext(os.path.basename(args.predict_file))[0])[0] 21 | print('{}\t{:.6f}'.format(model_name, AUC(y, p))) 22 | -------------------------------------------------------------------------------- /src/generate_j1.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | from __future__ import division 3 | import argparse 4 | import logging 5 | import numpy as np 6 | import os 7 | import pandas as pd 8 | import time 9 | 10 | from kaggler.data_io import load_data, save_data 11 | from kaggler.preprocessing import LabelEncoder 12 | 13 | from const import ID_COL, TARGET_COL 14 | 15 | 16 | def generate_feature(train_file, test_file, train_feature_file, 17 | test_feature_file, feature_map_file): 18 | logging.info('loading raw data') 19 | trn = pd.read_csv(train_file, index_col=ID_COL) 20 | tst = pd.read_csv(test_file, index_col=ID_COL) 21 | 22 | y = trn[TARGET_COL].values 23 | n_trn = trn.shape[0] 24 | 25 | trn.drop(TARGET_COL, axis=1, inplace=True) 26 | 27 | cat_cols = [x for x in trn.columns if trn[x].dtype == np.object] 28 | num_cols = [x for x in trn.columns if trn[x].dtype != np.object] 29 | 30 | logging.info('categorical: {}, numerical: {}'.format(len(cat_cols), 31 | len(num_cols))) 32 | 33 | df = pd.concat([trn, tst], axis=0) 34 | 35 | logging.info('label encoding categorical variables') 36 | lbe = LabelEncoder(min_obs=10) 37 | df[cat_cols] = lbe.fit_transform(df[cat_cols]) 38 | df[num_cols] = df[num_cols].fillna(-1) 39 | 40 | with open(feature_map_file, 'w') as f: 41 | for i, col in enumerate(df.columns): 42 | f.write('{}\t{}\tq\n'.format(i, col)) 43 | 44 | logging.info('saving features') 45 | save_data(df.values[:n_trn,], y, train_feature_file) 46 | save_data(df.values[n_trn:,], None, test_feature_file) 47 | 48 | 49 | if __name__ == '__main__': 50 | 51 | logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', 52 | level=logging.DEBUG) 53 | 54 | parser = argparse.ArgumentParser() 55 | parser.add_argument('--train-file', required=True, dest='train_file') 56 | parser.add_argument('--test-file', required=True, dest='test_file') 57 | parser.add_argument('--train-feature-file', required=True, dest='train_feature_file') 58 | parser.add_argument('--test-feature-file', required=True, dest='test_feature_file') 59 | parser.add_argument('--feature-map-file', required=True, dest='feature_map_file') 60 | 61 | args = parser.parse_args() 62 | 63 | start = time.time() 64 | generate_feature(args.train_file, 65 | args.test_file, 66 | args.train_feature_file, 67 | args.test_feature_file, 68 | args.feature_map_file) 69 | logging.info('finished ({:.2f} sec elasped)'.format(time.time() - start)) 70 | 71 | -------------------------------------------------------------------------------- /src/generate_n1.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | from __future__ import division 3 | from scipy import sparse 4 | import argparse 5 | import logging 6 | import numpy as np 7 | import os 8 | import pandas as pd 9 | import time 10 | 11 | from kaggler.data_io import load_data, save_data 12 | from kaggler.preprocessing import OneHotEncoder, Normalizer 13 | 14 | from const import ID_COL, TARGET_COL 15 | 16 | 17 | def generate_feature(train_file, test_file, train_feature_file, 18 | test_feature_file, feature_map_file): 19 | logging.info('loading raw data') 20 | trn = pd.read_csv(train_file, index_col=ID_COL) 21 | tst = pd.read_csv(test_file, index_col=ID_COL) 22 | 23 | y = trn[TARGET_COL].values 24 | n_trn = trn.shape[0] 25 | 26 | trn.drop(TARGET_COL, axis=1, inplace=True) 27 | 28 | cat_cols = [x for x in trn.columns if trn[x].dtype == np.object] 29 | num_cols = [x for x in trn.columns if trn[x].dtype != np.object] 30 | 31 | logging.info('categorical: {}, numerical: {}'.format(len(cat_cols), 32 | len(num_cols))) 33 | 34 | df = pd.concat([trn, tst], axis=0) 35 | 36 | logging.info('normalizing numeric features') 37 | nm = Normalizer() 38 | df[num_cols] = nm.fit_transform(df[num_cols].values) 39 | 40 | logging.info('label encoding categorical variables') 41 | ohe = OneHotEncoder(min_obs=10) 42 | X_ohe = ohe.fit_transform(df[cat_cols]) 43 | ohe_cols = ['ohe{}'.format(i) for i in range(X_ohe.shape[1])] 44 | 45 | X = sparse.hstack((df[num_cols].values, X_ohe), format='csr') 46 | 47 | with open(feature_map_file, 'w') as f: 48 | for i, col in enumerate(num_cols + ohe_cols): 49 | f.write('{}\t{}\tq\n'.format(i, col)) 50 | 51 | logging.info('saving features') 52 | save_data(X[:n_trn,], y, train_feature_file) 53 | save_data(X[n_trn:,], None, test_feature_file) 54 | 55 | 56 | if __name__ == '__main__': 57 | 58 | logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', 59 | level=logging.DEBUG) 60 | 61 | parser = argparse.ArgumentParser() 62 | parser.add_argument('--train-file', required=True, dest='train_file') 63 | parser.add_argument('--test-file', required=True, dest='test_file') 64 | parser.add_argument('--train-feature-file', required=True, dest='train_feature_file') 65 | parser.add_argument('--test-feature-file', required=True, dest='test_feature_file') 66 | parser.add_argument('--feature-map-file', required=True, dest='feature_map_file') 67 | 68 | args = parser.parse_args() 69 | 70 | start = time.time() 71 | generate_feature(args.train_file, 72 | args.test_file, 73 | args.train_feature_file, 74 | args.test_feature_file, 75 | args.feature_map_file) 76 | logging.info('finished ({:.2f} sec elasped)'.format(time.time() - start)) 77 | 78 | -------------------------------------------------------------------------------- /src/train_predict_krs1.py: -------------------------------------------------------------------------------- 1 | from keras.callbacks import EarlyStopping 2 | from keras.models import Sequential 3 | from keras.layers import Dense, Dropout, Activation 4 | from keras.layers.normalization import BatchNormalization 5 | from keras.layers.advanced_activations import PReLU 6 | from keras.utils import np_utils, generic_utils 7 | from scipy.sparse import csr_matrix, hstack 8 | from sklearn.preprocessing import LabelEncoder, OneHotEncoder, StandardScaler 9 | from sklearn.model_selection import StratifiedKFold 10 | from sklearn.metrics import roc_auc_score as AUC 11 | 12 | import argparse 13 | import logging 14 | import keras.backend as K 15 | import numpy as np 16 | import os 17 | import pandas as pd 18 | import time 19 | 20 | 21 | from kaggler.data_io import load_data 22 | from const import N_FOLD, SEED 23 | 24 | 25 | np.random.seed(SEED) # for reproducibility 26 | 27 | 28 | def batch_generator(X, y, batch_size, shuffle): 29 | number_of_batches = np.ceil(X.shape[0]/batch_size) 30 | counter = 0 31 | sample_index = np.arange(X.shape[0]) 32 | if shuffle: 33 | np.random.shuffle(sample_index) 34 | while True: 35 | batch_index = sample_index[batch_size*counter:batch_size*(counter+1)] 36 | X_batch = X[batch_index,:].toarray() 37 | y_batch = y[batch_index] 38 | counter += 1 39 | yield X_batch, y_batch 40 | if (counter == number_of_batches): 41 | if shuffle: 42 | np.random.shuffle(sample_index) 43 | counter = 0 44 | 45 | 46 | def batch_generatorp(X, batch_size, shuffle): 47 | number_of_batches = X.shape[0] / np.ceil(X.shape[0]/batch_size) 48 | counter = 0 49 | sample_index = np.arange(X.shape[0]) 50 | while True: 51 | batch_index = sample_index[batch_size * counter:batch_size * (counter + 1)] 52 | X_batch = X[batch_index, :].toarray() 53 | counter += 1 54 | yield X_batch 55 | if (counter == number_of_batches): 56 | counter = 0 57 | 58 | 59 | def nn_model(dims): 60 | model = Sequential() 61 | 62 | model.add(Dense(400, input_dim=dims, kernel_initializer='he_normal')) 63 | model.add(PReLU()) 64 | model.add(BatchNormalization()) 65 | model.add(Dropout(0.4)) 66 | 67 | model.add(Dense(200, kernel_initializer='he_normal')) 68 | model.add(PReLU()) 69 | model.add(BatchNormalization()) 70 | model.add(Dropout(0.2)) 71 | 72 | model.add(Dense(50, kernel_initializer='he_normal')) 73 | model.add(PReLU()) 74 | model.add(BatchNormalization()) 75 | model.add(Dropout(0.2)) 76 | 77 | model.add(Dense(1, kernel_initializer='he_normal', activation='sigmoid')) 78 | model.compile(loss = 'binary_crossentropy', optimizer = 'adadelta') 79 | return(model) 80 | 81 | 82 | def train_predict(train_file, test_file, predict_valid_file, predict_test_file, 83 | n_est=100, batch_size=1024, retrain=True): 84 | 85 | model_name = os.path.splitext(os.path.splitext(os.path.basename(predict_test_file))[0])[0] 86 | 87 | logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', 88 | level=logging.DEBUG, 89 | filename='{}.log'.format(model_name)) 90 | 91 | logging.info('Loading training and test data...') 92 | X, y = load_data(train_file) 93 | X_tst, _ = load_data(test_file) 94 | 95 | dims = X.shape[1] 96 | logging.info('{} dims'.format(dims)) 97 | 98 | logging.info('Loading CV Ids') 99 | cv = StratifiedKFold(n_splits=N_FOLD, shuffle=True, random_state=SEED) 100 | 101 | p = np.zeros_like(y) 102 | p_tst = np.zeros((X_tst.shape[0],)) 103 | for i, (i_trn, i_val) in enumerate(cv.split(X, y), 1): 104 | logging.info('Training model #{}'.format(i)) 105 | clf = nn_model(dims) 106 | clf.fit_generator(generator=batch_generator(X[i_trn], 107 | y[i_trn], 108 | batch_size, 109 | True), 110 | nb_epoch=n_est, 111 | samples_per_epoch=X[i_trn].shape[0], 112 | verbose=0) 113 | 114 | p[i_val] = clf.predict_generator(generator=batch_generatorp(X[i_val], batch_size, False), 115 | val_samples=X[i_val].shape[0])[:, 0] 116 | logging.info('CV #{}: {:.4f}'.format(i, AUC(y[i_val], p[i_val]))) 117 | 118 | if not retrain: 119 | p_tst += clf.predict_generator(generator=batch_generatorp(X_tst, batch_size, False), 120 | val_samples=X_tst.shape[0])[:, 0] / N_FOLD 121 | 122 | logging.info('Saving validation predictions...') 123 | logging.info('CV: {:.4f}'.format(AUC(y, p))) 124 | np.savetxt(predict_valid_file, p, fmt='%.6f', delimiter=',') 125 | 126 | if retrain: 127 | logging.info('Retraining with 100% training data') 128 | clf = nn_model(dims) 129 | clf.fit_generator(generator=batch_generator(X, Y, batch_size, True), 130 | nb_epoch=n_est) 131 | p_tst = clf.predict_generator(generator=batch_generatorp(X_tst, batch_size, False), 132 | val_samples=X_tst.shape[0])[:, 0] 133 | 134 | logging.info('Saving normalized test predictions...') 135 | np.savetxt(predict_test_file, p_tst, fmt='%.6f', delimiter=',') 136 | 137 | 138 | if __name__ == '__main__': 139 | parser = argparse.ArgumentParser() 140 | parser.add_argument('--train-file', required=True, dest='train_file') 141 | parser.add_argument('--test-file', required=True, dest='test_file') 142 | parser.add_argument('--predict-valid-file', required=True, 143 | dest='predict_valid_file') 144 | parser.add_argument('--predict-test-file', required=True, 145 | dest='predict_test_file') 146 | parser.add_argument('--n-est', default=10, type=int, dest='n_est') 147 | parser.add_argument('--batch-size', default=64, type=int, 148 | dest='batch_size') 149 | parser.add_argument('--hiddens', default=2, type=int) 150 | parser.add_argument('--neurons', default=512, type=int) 151 | parser.add_argument('--dropout', default=0.5, type=float) 152 | parser.add_argument('--retrain', default=False, action='store_true') 153 | 154 | args = parser.parse_args() 155 | 156 | start = time.time() 157 | train_predict(train_file=args.train_file, 158 | test_file=args.test_file, 159 | predict_valid_file=args.predict_valid_file, 160 | predict_test_file=args.predict_test_file, 161 | n_est=args.n_est, 162 | batch_size=args.batch_size, 163 | retrain=args.retrain) 164 | logging.info('finished ({:.2f} min elasped)'.format((time.time() - start) / 165 | 60.)) 166 | -------------------------------------------------------------------------------- /src/train_predict_lgb1.py: -------------------------------------------------------------------------------- 1 | #!/usr/bin/env python 2 | 3 | from sklearn.model_selection import KFold, StratifiedKFold 4 | from sklearn.metrics import roc_auc_score as AUC 5 | 6 | import argparse 7 | import logging 8 | import numpy as np 9 | import operator 10 | import os 11 | import pandas as pd 12 | import time 13 | 14 | from const import N_FOLD, SEED 15 | from kaggler.data_io import load_data 16 | 17 | import lightgbm as lgb 18 | 19 | 20 | def train_predict(train_file, test_file, predict_valid_file, predict_test_file, 21 | n_est=100, n_leaf=200, lrate=.1, n_min=8, subcol=.3, subrow=.8, 22 | subrow_freq=100, n_stop=100, retrain=True): 23 | 24 | model_name = os.path.splitext(os.path.splitext(os.path.basename(predict_test_file))[0])[0] 25 | 26 | logging.basicConfig(format='%(asctime)s %(levelname)s %(message)s', 27 | level=logging.DEBUG, 28 | filename='{}.log'.format(model_name)) 29 | 30 | logging.info('Loading training and test data...') 31 | X, y = load_data(train_file) 32 | X_tst, _ = load_data(test_file) 33 | 34 | logging.info('Loading CV Ids') 35 | cv = StratifiedKFold(n_splits=N_FOLD, shuffle=True, random_state=SEED) 36 | 37 | params = {'random_state': SEED, 38 | 'n_jobs': -1, 39 | 'objective': 'binary', 40 | 'boosting': 'gbdt', 41 | 'learning_rate': lrate, 42 | 'num_leaves': n_leaf, 43 | 'feature_fraction': subcol, 44 | 'bagging_fraction': subrow, 45 | 'bagging_freq': subrow_freq, 46 | 'verbosity': -1, 47 | 'min_child_samples': n_min, 48 | 'metric': 'auc'} 49 | 50 | p = np.zeros(X.shape[0]) 51 | p_tst = np.zeros(X_tst.shape[0]) 52 | n_bests = [] 53 | for i, (i_trn, i_val) in enumerate(cv.split(X, y), 1): 54 | logging.info('Training model #{}'.format(i)) 55 | trn_lgb = lgb.Dataset(X[i_trn], label=y[i_trn]) 56 | val_lgb = lgb.Dataset(X[i_val], label=y[i_val]) 57 | 58 | logging.info('Training with early stopping') 59 | clf = lgb.train(params, trn_lgb, n_est, val_lgb, early_stopping_rounds=n_stop, verbose_eval=100) 60 | n_best = clf.best_iteration 61 | n_bests.append(n_best) 62 | logging.info('best iteration={}'.format(n_best)) 63 | 64 | p[i_val] = clf.predict(X[i_val]) 65 | logging.info('CV #{}: {:.4f}'.format(i, AUC(y[i_val], p[i_val]))) 66 | 67 | if not retrain: 68 | p_tst += clf.predict(X_tst) / N_FOLD 69 | 70 | logging.info('CV: {:.4f}'.format(AUC(y, p))) 71 | logging.info('Saving validation predictions...') 72 | np.savetxt(predict_valid_file, p, fmt='%.6f', delimiter=',') 73 | 74 | if retrain: 75 | logging.info('Retraining with 100% training data') 76 | n_best = sum(n_bests) // N_FOLD 77 | clf = lgb.LGBMRegressor(n_estimators=n_best, 78 | num_leaves=n_leaf, 79 | learning_rate=lrate, 80 | min_child_samples=n_min, 81 | subsample=subrow, 82 | subsample_freq=subrow_freq, 83 | colsample_bytree=subcol, 84 | objective=fairobj, 85 | nthread=1, 86 | seed=SEED) 87 | 88 | clf = clf.fit(X, y) 89 | p_tst = clf.predict(X_tst) 90 | 91 | logging.info('Saving test predictions...') 92 | np.savetxt(predict_test_file, p_tst, fmt='%.6f', delimiter=',') 93 | 94 | 95 | if __name__ == '__main__': 96 | parser = argparse.ArgumentParser() 97 | parser.add_argument('--train-file', required=True, dest='train_file') 98 | parser.add_argument('--test-file', required=True, dest='test_file') 99 | parser.add_argument('--predict-valid-file', required=True, 100 | dest='predict_valid_file') 101 | parser.add_argument('--predict-test-file', required=True, 102 | dest='predict_test_file') 103 | parser.add_argument('--n-est', type=int, dest='n_est') 104 | parser.add_argument('--n-leaf', type=int, dest='n_leaf') 105 | parser.add_argument('--lrate', type=float) 106 | parser.add_argument('--subcol', type=float, default=1) 107 | parser.add_argument('--subrow', type=float, default=.5) 108 | parser.add_argument('--subrow-freq', type=int, default=100, 109 | dest='subrow_freq') 110 | parser.add_argument('--n-min', type=int, default=1, dest='n_min') 111 | parser.add_argument('--early-stop', type=int, dest='n_stop') 112 | parser.add_argument('--retrain', default=False, action='store_true') 113 | 114 | args = parser.parse_args() 115 | 116 | start = time.time() 117 | train_predict(train_file=args.train_file, 118 | test_file=args.test_file, 119 | predict_valid_file=args.predict_valid_file, 120 | predict_test_file=args.predict_test_file, 121 | n_est=args.n_est, 122 | n_leaf=args.n_leaf, 123 | lrate=args.lrate, 124 | n_min=args.n_min, 125 | subcol=args.subcol, 126 | subrow=args.subrow, 127 | subrow_freq=args.subrow_freq, 128 | n_stop=args.n_stop, 129 | retrain=args.retrain) 130 | logging.info('finished ({:.2f} min elasped)'.format((time.time() - start) / 131 | 60)) 132 | -------------------------------------------------------------------------------- /tox.ini: -------------------------------------------------------------------------------- 1 | [tox] 2 | envlist = py36 3 | 4 | [testenv] 5 | setenv = 6 | PYTHONPATH = {toxinidir} 7 | deps = pytest 8 | commands = 9 | pytest -sv tests/ --cov={toxinidir}/kaggler 10 | - codecov 11 | 12 | [flake8] 13 | max-line-length = 120 14 | ignore = D001, # lines should not be longer than 79 characters, plus some exceptions 15 | E121, # Continuation line under-indented for hanging indent 16 | E123, # Closing bracket does not match indentation of opening bracket's line 17 | E126, # Continuation line over-indented for hanging indent 18 | E128, # Continuation line under-indented for visual indent 19 | E129, # Visually indented line with same indent as next logical line 20 | E226, # Missing whitespace around arithmetic operator 21 | E24, # 22 | E704, # Multiple statements on one line (def) 23 | E731, # Do not assign a lambda expression, use a def 24 | E741, # Ambiguous variable name 25 | W503 # Line break before binary operator 26 | W504 # Line break after binary operator 27 | 28 | builtins = __builtins__ 29 | --------------------------------------------------------------------------------